Preliminary steps

Load necessary libraries and files

Load libraries

Load files

Mean dominance curves by technique

Clusters for training set

Get the clusters for K = 6

Make sure data is numeric

Get the clusters

Obtain the dataset kappa mean for each cluster

Unify into one dataframe

Plot the results

# Get Kappa loss in order to print Kappa Loss Curves
# Calculate 1 - mean_kappa rounded to two decimals
mean_results$kappa_loss <- round(1 - mean_results$mean_kappa, 3)

Noise 10

for(t in techniques){
      # Filter data for the current dataset and method
      filtered_data <- subset(mean_results, technique == t & noise == 10)
      
      # Create plot
      p <- ggplot(filtered_data, aes(x = percentage, y = kappa_loss, color = factor(cluster))) +
        geom_point() +
        geom_line(aes(group = factor(cluster))) +
        labs(x = "Percentage of altered instances", y = bquote(.(nabla ~ "Kappa")), " Kappa", color = "Cluster") +
        ggtitle(paste("Technique: ", t, " Noise injected: ", 10)) +
        theme_bw() +
        scale_y_continuous(limits = c(0.0, 1), breaks = seq(0, 1, by = 0.1))
      
      # Store plot in the list
      plot_name <- paste(t, 10, sep = "_")
      plot_list[[plot_name]] <- p
      print(p)
}

Noise 20

for(t in techniques){
      # Filter data for the current dataset and method
      filtered_data <- subset(mean_results, technique == t & noise == 20)
      
      # Create plot
      p <- ggplot(filtered_data, aes(x = percentage, y = kappa_loss, color = factor(cluster))) +
        geom_point() +
        geom_line(aes(group = factor(cluster))) +
        labs(x = "Percentage of altered instances", y = bquote(.(nabla ~ "Kappa")), " Kappa", color = "Cluster") +
        ggtitle(paste("Technique: ", t, " Noise injected: ", 20)) +
        theme_bw() +
        scale_y_continuous(limits = c(0.0, 1), breaks = seq(0, 1, by = 0.1))
      
      # Store plot in the list
      plot_name <- paste(t, 20, sep = "_")
      plot_list[[plot_name]] <- p
      print(p)
}

Noise 30

for(t in techniques){
      # Filter data for the current dataset and method
      filtered_data <- subset(mean_results, technique == t & noise == 30)
      
      # Create plot
      p <- ggplot(filtered_data, aes(x = percentage, y = kappa_loss, color = factor(cluster))) +
        geom_point() +
        geom_line(aes(group = factor(cluster))) +
        labs(x = "Percentage of altered instances", y = bquote(.(nabla ~ "Kappa")), " Kappa", color = "Cluster") +
        ggtitle(paste("Technique: ", t, " Noise injected: ", 30)) +
        theme_bw() +
        scale_y_continuous(limits = c(0.0, 1), breaks = seq(0, 1, by = 0.1))
      
      # Store plot in the list
      plot_name <- paste(t, 30, sep = "_")
      plot_list[[plot_name]] <- p
      print(p)
}

Save all plots